My AI Setup

AI setup for product work, split into installable plugins plus a flat browsing layer.

This repository follows the same structural idea as Basecamp's house-skills: real skill files live inside their owning plugin, while top-level skills/ contains symlinks for a unified view.

Install

Install from the marketplace, then choose the plugin you need:

/plugin marketplace add xu-chris/ai-setup
/plugin install shape-up@ai-setup
/plugin install interface-design@ai-setup
/plugin install frontend@ai-setup
/plugin install ai@ai-setup
/plugin install security@ai-setup
/plugin install elixir@ai-setup
/plugin install tools@ai-setup
/plugin install delivery@ai-setup

Other agents can install the standalone skills view:

npx skills add xu-chris/ai-setup

Use shape-up for dedicated project setup. Use elixir for Elixir/Phoenix/Ash/BEAM projects. Use delivery when Basecamp cards are the work queue. Use interface-design, frontend, ai, security, and tools as user-level setups when you want those capabilities available across projects.

Plugins

Plugin	Use	Contents
shape-up	Dedicated project workflow	Strategy facilitation, frame, shape, bet, kickoff, scope, breadboarding, breadboard reflection, DAG
interface-design	User-level design setup	UX design and usability audit skill
frontend	User-level frontend setup	Semantic HTML, CSS architecture, layout, accessibility, component patterns
ai	AI setup authoring	Skill crafting and install documentation
security	CI security setup	GitHub Actions hardening
elixir	Elixir/Phoenix/Ash/BEAM setup	Architecture, Ash code, BEAM security, cleanup, dependency updates
tools	Tool operation setup	Firecrawl web search, scraping, research, content extraction
delivery	Card delivery setup	Basecamp card implementation, expert code review, worktree isolation, verified commits, PR handoff

Skills By Plugin

shape-up

Skill	Description
facilitating-strategy	Develop, sharpen, and pressure-test real strategy before shaping work
frame	Turn a raw idea, request, or complaint into a clear problem frame
shape	Define a solution at Shape Up abstraction: rough, solved, bounded
breadboarding	Prototype the wiring: places, affordances, stores, flows
breadboard-reflection	Sync breadboards to implementation and inspect design smells
bet	Decide whether shaped work deserves protected build time
kickoff	Move a bet into build by mapping vertical scopes
scope	Break one scope into concrete tasks at the start of that scope
dag	Render scope or task dependency graphs

interface-design

Skill	Description
ux-design	Design UI behavior, copy, errors, empty states, and audits

frontend

Skill	Description
frontend-craft	Write and review semantic, accessible HTML and maintainable CSS

ai

Skill	Description
skill-crafting	Create, edit, and refine agent skills through co-development and eval loops
install-md	Create installation docs optimized for autonomous agent execution

security

Skill	Description
harden-github-actions	Resolve zizmor warnings and harden GitHub Actions workflows

elixir

Skill	Description
elixir-architect	Design Elixir/Phoenix/Ash architecture and handoff documentation
writing-ash-code	Write production-grade Ash resources, actions, policies, forms, and jobs
beam-security-hardening	Harden BEAM/Phoenix code, deployments, LiveView, and Zero Trust decisions
codebase-cleanup	Run focused BEAM cleanup and remediation passes
updating-dependencies	Update Elixir/Phoenix dependencies safely

tools

Skill	Description
firecrawl	Use Firecrawl CLI for web search, scraping, research, and extraction

delivery

Skill	Description
programming-a-card	Take a Basecamp card from ready to implemented with evidence, review, PR, and card updates
code-review-expert	Review diffs with senior-engineer rigor for bugs, architecture, security, tests, and removal candidates

Commands

Top-level commands/ mirrors plugin commands with symlinks for browsing.

Command	Plugin	What it does
/shape-up:frame	shape-up	Examine a candidate idea and produce a problem statement
/shape-up:shape	shape-up	Define solution elements, breadboard, rabbit holes, dos, and won't-dos
/shape-up:bet	shape-up	Decide whether to commit build time
/shape-up:kickoff	shape-up	Tour the concept and map vertical scopes
/shape-up:scope	shape-up	Break one scope into concrete tasks
/shape-up:dag	shape-up	Generate a dependency graph for scopes or tasks

Structure

plugins/
  shape-up/
    .claude-plugin/plugin.json
    commands/
    hooks/
    skills/
  interface-design/
    .claude-plugin/plugin.json
    skills/
  frontend/
    .claude-plugin/plugin.json
    skills/
  ai/
    .claude-plugin/plugin.json
    skills/
  security/
    .claude-plugin/plugin.json
    skills/
  elixir/
    .claude-plugin/plugin.json
    skills/
  tools/
    .claude-plugin/plugin.json
    skills/
  delivery/
    .claude-plugin/plugin.json
    skills/
skills/      # symlinks to plugins/{plugin}/skills/{skill}
commands/    # symlinks to plugins/shape-up/commands/{command}.md
scripts/
bin/
tests/

Real files belong inside plugins/{plugin}/. The top-level skills/ directory must contain only symlinks. This keeps plugin extraction self-contained and still gives agents and humans one flat skill index.

Run validation before publishing:

bin/ci

Hooks

shape-up includes a ripple-check hook. It fires when a Markdown file carries shaping: true in the first five lines of frontmatter and reminds the agent to keep downstream shaping documents consistent:

• Breadboard diagram changed: update affordance tables first, then re-render Mermaid.
• Requirements in shape.md changed: update the Fit Check and slices.md when needed.
• Shape parts changed: update the Fit Check.
• slices.md changed: update affected S#-plan.md files.
• S#-plan.md changed: update the slice summary in slices.md if scope changed.

Sources

• Shape Up
• rjs/shaping-skills
• obra/superpowers
• basecamp/house-skills

MIT License